LlamaIndex

Log inLog into community

Find answers from the community

Updated last year

Hey ll

Hey ll

AAtaliba Miguel

·

Hey'll,

I am really struggling here, does anyone could provide support in guiding me on how/where to find the info under the mongoDB account? Some screenshots on how/where to find the required info for: (It is related to mongodb loader)

host = ?????
port = ?????

db_name = MONGODB_DATABASE
collection_name = ???? # is this the same name for collections? "Collections"

query_dict is passed into db.collection.find()

query_dict = {}

reader = SimpleMongoReader(host, port)

documents = reader.load_data(
db_name, collection_name, query_dict=query_dict)

Attachments

c

V

A

27 comments

When you're accessing this database, what host/port do you use? For example, to load the data into it

Did you created the index one by one with hand

or with the llamaindex program ?

AAtaliba Miguel

Hi @Vaylonn the index was created with llamaindex, i.e., I have used PDFReader to load the to storage on mongoDB. Now I plan only to access files stored under mongoDB for query without resorting loading them from memory. I am kind of trying to test question-retrieval directly from the files in the Mongo_Database.

AAtaliba Miguel

Hi @chris, I am accessing from mongodb+srv://apcmiguel:**@cluster0.dkehzhx.mongodb.net/

AAtaliba Miguel

Attachment

it looks like you've included the port in the host string. Try removing that?

for the collection name, you'll need to specify one of the collections in the database. I'm not familiar with this UI but you could try docstore/data

From what I did, I created index under different names

Plain Text

mongodb_client = pymongo.MongoClient(mongo_uri)
    db_name = f"{dossier}"
    store = MongoDBAtlasVectorSearch(mongodb_client, db_name=db_name)


    # Création ou mise à jour d'un index à partir de documents dans le dossier 'Sources'
    # A commenter/décommenter si on veut créer ou mettre à jour un index

    storage_context = StorageContext.from_defaults(vector_store=store)
    docs = SimpleDirectoryReader("./Sources").load_data()
    index = VectorStoreIndex.from_documents(docs, storage_context=storage_context)

by specifying db_name

and i can load different indexes like that

don't forget that without a search engine created in mongoDB you will not have the possibilty to query over the index

I didnt bother with ports

oooooh

maybe you mean that you can't access any index at all

because you can't enter the mongo network

AAtaliba Miguel

an Hi @Vaylonn, could you please indicate from the connection string (in the attached img) where to find the Host and the Port? I am still blind and unable to grasp it.
I tried your code but got lost on the MongoDBAtlasVectorSearch(mongodb_client, db_name=db_name), it was giving error.

So I am persisting on the code snippet as below:

if submit_button:
num_output = 512

llm_predictor = LLMPredictor(llm=ChatOpenAI(temperature=0, model_name="gpt-3.5-turbo", max_tokens=num_output))

define LLM service context

service_context = ServiceContext.from_defaults(llm_predictor=llm_predictor, chunk_size=1024)

MONGO_URI = os.environ["MONGO_URI"]
MONGODB_DATABASE = "AI"

SimpleMongoReader = download_loader('SimpleMongoReader')

host = "<host>"
port = "<port>"
db_name = "<db_name>"
collection_name = "<collection_name>"

query_dict is passed into db.collection.find()

query_dict = {}

reader = SimpleMongoReader(host, port)

documents = reader.load_data(db_name, collection_name, query_dict=query_dict)

nodes = service_context.node_parser.get_nodes_from_documents(documents)

storage_context = StorageContext.from_defaults(docstore=MongoDocumentStore.from_uri(uri=MONGO_URI, db_name=MONGODB_DATABASE),
index_store=MongoIndexStore.from_uri(uri=MONGO_URI, db_name=MONGODB_DATABASE),)

list_index = GPTListIndex(nodes, storage_context=storage_context)
vector_index = GPTVectorStoreIndex.from_documents(nodes, storage_context=storage_context)
keyword_table_index = GPTSimpleKeywordTableIndex(nodes, storage_context=storage_context)

vector_response = vector_index.as_query_engine().query(question)
st.success(vector_response)

Attachment

actually i didnt get trough this way since i'm using Azure open AI

but i did that

with mongodbatlas

Plain Text

mongo_uri = os.environ["MONGO_URI"] = "mongodb+srv://sxdinno:xxxx@indexatlasdb.vwrbmy3.mongodb.net/index?retryWrites=true&w=majority"

some code

then this only

Plain Text

mongodb_client = pymongo.MongoClient(mongo_uri)
    db_name = f"{dossier}"
    store = MongoDBAtlasVectorSearch(mongodb_client, db_name=db_name)


    # Création ou mise à jour d'un index à partir de documents dans le dossier 'Sources'
    # A commenter/décommenter si on veut créer ou mettre à jour un index

    # storage_context = StorageContext.from_defaults(vector_store=store)
    # docs = SimpleDirectoryReader("./Sources").load_data()
    # index = VectorStoreIndex.from_documents(docs, storage_context=storage_context)


    # Initialisation de l'index via les index sur MongoDB Atlas
    # Et inversement, commenter/décommenter si on veut juste query l'index existant

    index = VectorStoreIndex.from_vector_store(store)

https://discord.com/channels/1059199217496772688/1122793802214670336/1123217023732817940

whatch this thread

Add a reply

Sign up and join the conversation on Discord